In the fields of packet communication voice storage typified by digital radio communication and Internet communication, coding/decoding technology for speech signals is indispensable for effective utilization of transmission path capacity for radio waves and recording media, and many speech coding/decoding schemes have been developed so far.
At present, CELP-based speech coding/decoding schemes are used in mainstream (e.g., Non-Patent Document 1). A CELP-based speech coding scheme is designed, primarily, to store models of vocalized sounds and encode input speech based on prestored voice models.
In recent years, in coding of speech signals and audio signals, a scalable coding technology using the CELP scheme is being developed, which is capable of decoding speech/audio sound signals even from part of coded information and minimizing quality deterioration even in the situation where packet loss occurs.
According to a typical example of this scalable coding technology, a method is known whereby an input signal is coded by a coding section of a first layer and coded information is acquired. In addition, for an upper layer, a residual signal, which is the difference between an input signal of a coding section of the (i−1)th layer (i: an integer of two or greater), and a decoded signal of coded information of the (i−1)th layer is used as an input signal for the ith layer, and this signal is coded by a coding section of the ith layer and coded information is acquired, and these processes are repeated.
A decoding apparatus that decodes coded information coded using this method performs decoding in the reverse order of coding.
A decoding apparatus using a conventional scalable coding technology employs a method, whereby, when packet loss occurs (when error is detected) in a layer, a decoded signal of the same layer is processed as an unvoiced signal, or decoding is performed using a decoding result of a packet in at least one of the immediately preceding coded frame and the immediately following coded frame in the same layer (e.g., Patent Document 1).
Patent Document 1: Japanese Patent Application Laid-open No.2003-241799.
Non-Patent Document 1: M. R. Schroeder, B. S. Atal, “Code Excited Linear Prediction: High Quality Speech at Very Low Bit Rate,” IEEE proc., ICASSP'85 pp. 937-940.